AI techniques can be used in a variety of ways within technology that enables people with cognitive impairment to live more autonomously.    We have employed automated planning, temporal reasoning, machine learning, and probabilistic inference techniques to design and investigate systems that assist cognitively impaired people by providing them with flexibly timed reminders of daily activities; we are also exploring systems that can do continuous, naturalistic assessment of functional performance.

Selected Papers:

M. R. Hodges and M. E. Pollack, "An ‘Object-Use Fingerprint’:  The Use of Electronic Sensors for Human Identification,” 9^th International Conference on Ubiquitous Computing, Sept. 2007.

M. E. Pollack, "Intelligent Technology for an Aging Population:  The Use of AI to Assist Elders with Cognitive Impairment,"  AI Magazine, 26(2):9-24, 2005.

M. Rudary, S. Singh, and M. E. Pollack, "Adaptive Cognitive Orthotics:  Combining Reinforcement Learning and Constraint-Based Temporal Reasoning,"  21st International Conference on Machine Learning, July 2004.

This work has primarily been funded by the NSF and the Intel Corp.

Constraint-Based Temporal Reasoning

Many important applications involve reasoning about time.   Constraint-satisfaction processing provides a flexible and powerful framework for formalizing temporal reasoning.  We study richly expressive models that permit the representation of disjunction, temporal and causal uncertainty, soft constraints (preferences), and hybrid (temporal and finite-domain) constraints and we develop efficient algorithms for performing inference with them.

Selected Papers:

M. D. Moffitt and M. E. Pollack, "Generalizing Temporal Controllability," Proceedings of the 20th International Joint Conference on Artificial Intelligence, Jan. 2007.

M. D. Moffitt and M. E. Pollack, "Temporal Preference Optimization as Weighted Constraint Satisfaction,"   Proceedings of the 21st National Conference on Artificial Intelligence, July 2006.

H. Sheini, B. Peintner, K. Sakallah, and M. E. Pollack, “On Solving Soft Temporal Constraints using SAT Techniques,” Proceedings of the 11th International Conference on Principles and Practice of Constraint Programming, Oct. 2005.

This work has primarily been funded by AFOSR, DARPA, and the NSF.

We are applying machine learning and constraint-satisfaction techniques to develop computer interfaces that adapt to the needs and preferences of their users.  A central challenge arises from the fact that in an interactive system, training examples cannot constructed arbitrarily; instead, naturally occurring interactions must be exploited to balance learning convergence speed with user satisfaction.

Selected Papers:

J. S. Weber and M. E. Pollack, "Entropy-Driven Online Active Learning for Interactive Calendar Management," Proceedings of the 10th International Conference on Intelligent User Interfaces, Jan. 2007.

K. Myers, P. Berry, J. Blythe, K. Conley, M. Gervasio, D. McGuinness, D. Morley, A. Pfeffer, M. Pollack, and M. Tambe, "An Intelligent Personal Assistant for Task and Time Management,"  AI Magazine, 2007.

M. T. Gervasio, M. D. Moffitt, M. E. Pollack, J. M. Taylor, and T. E. Uribe, "Active Preference Learning for Personalized Calendar Scheduling Assistance,"  International Conference on Intelligent User Interfaces, January 2005.

This work has primarily been funded by DARPA.

In the past, I have worked on a variety of other topics in Artificial Intelligence, including:

• Discourse analysis for natural-language processing
• Computational models of rationality (BDI models)
• Automated plan generation and execution
• Software testing using planning models

See my publications for more information.